Schur-Weyl duality

Intuition [Beginner]

Schur-Weyl duality is one of the most elegant facts in representation theory. Take an $n$-fold tensor product of a vector space — the space whose elements are $n$-fold formal products of vectors, written $V^{(n)}$ for now (the standard symbol is introduced at the Intermediate tier). Two groups act on this space at the same time: the general linear group $\mathrm{GL}(V)$ acts on every factor simultaneously (the diagonal action), and the symmetric group $S_n$ acts by permuting the $n$ factors. These two actions commute. Schur (1901) and Weyl (1939) discovered that the two actions are not just compatible — they are dual in a very strong sense.

The duality says the $n$-fold tensor power breaks into a beautifully clean direct sum, indexed by partitions $\lambda$ of $n$. Each summand is the tensor product of an irreducible representation of $\mathrm{GL}(V)$ paired with an irreducible representation of the symmetric group. There is no overlap and no waste: every partition contributes exactly one piece, and the matching is bijective.

The reason this matters: it gives a way to read off everything you might want to know about $\mathrm{GL}(V)$ from the symmetric group, and vice versa. The character theory of $S_n$ — which is purely combinatorial and lives on partitions — controls the character theory of $\mathrm{GL}(V)$. Two areas of mathematics that look completely separate at first inspection are revealed to be different sides of one coin.

Visual [Beginner]

A picture of the tensor square of $V$ (the simplest case, $n=2$) splitting into two pieces: the symmetric square ${\mathrm{Sym}}^{2}V$ and the alternating square ${\Lambda }^{2}V$. The first is invariant under swapping factors; the second changes sign. The swap-by-$+1$ piece is paired with the unit representation of $S_2$, and the swap-by-$-1$ piece is paired with the sign representation. These are the two irreducible representations of $S_2$, and they correspond to the two partitions of $2$: the row partition $(2)$ and the column partition $(1,1)$.

Worked example [Beginner]

Work the case $n=2$ explicitly. Take $V={\mathbb{C}}^2$, the simplest non-zero vector space where the picture is rich. The tensor square of $V$ is a 4-dimensional space spanned by four formal products: take the two basis vectors $e_1,e_2$ and form the four formal products $(e_1,e_1)$, $(e_1,e_2)$, $(e_2,e_1)$, $(e_2,e_2)$ where the first entry sits in the first tensor factor and the second entry sits in the second.

Step 1. The symmetric group $S_2$ has two elements: the identity and the swap $\tau$. The swap sends the formal product $(e_i,e_j)$ to $(e_j,e_i)$. So $\tau$ fixes $(e_1,e_1)$ and $(e_2,e_2)$, and swaps $(e_1,e_2)$ with $(e_2,e_1)$.

Step 2. Decompose by the eigenspaces of $\tau$. The vectors fixed by $\tau$ are those with eigenvalue $+1$: a basis is $(e_1,e_1)$, $(e_2,e_2)$, and the symmetric combination $(e_1,e_2)+(e_2,e_1)$. This is a 3-dimensional space, the symmetric square ${\mathrm{Sym}}^{2}V$.

Step 3. The vectors flipped in sign by $\tau$ are those with eigenvalue $-1$: a basis is the antisymmetric combination $(e_1,e_2)-(e_2,e_1)$. This is 1-dimensional, the alternating square ${\Lambda }^{2}V$.

Step 4. Now check the matching with $\mathrm{GL}(V)$. The general linear group $\mathrm{GL}(2,\mathbb{C})$ acts on the tensor square by acting on every factor simultaneously: $g$ sends the formal product $(v,w)$ to $(gv,gw)$. This action commutes with the swap $\tau$ because $\tau$ does not see which factor is "first". So ${\mathrm{Sym}}^{2}V$ and ${\Lambda }^{2}V$ are both stable under $\mathrm{GL}(2,\mathbb{C})$, and each is an irreducible $\mathrm{GL}(2,\mathbb{C})$-representation. The 3-dimensional symmetric square corresponds to the partition $\lambda =(2)$; the 1-dimensional alternating square (the determinant representation) corresponds to $\lambda =(1,1)$.

Step 5. Read off Schur-Weyl: the tensor square of $V$ is the direct sum of two pieces. One piece is the symmetric square paired with the unit representation of $S_2$; the other is the alternating square paired with the sign representation of $S_2$. Each partition $\lambda$ of $2$ contributes one summand. The dimensions check: $3\cdot 1+1\cdot 1=4$, the dimension of the tensor square.

What this tells us: the two ways the tensor square can break up — by the swap action of $S_2$ on one side, and by the action of $\mathrm{GL}(V)$ on the other — are matched to each other exactly. The matching is indexed by partitions of $n=2$. The same pattern works for any $n$, and that is the content of Schur-Weyl duality.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Fix a finite-dimensional complex vector space $V$ of dimension $d$, and an integer $n\ge 1$. The $n$-fold tensor power $V^{\otimes n}$ carries two commuting actions:

Definition (diagonal $\mathrm{GL}(V)$-action). For $g\in \mathrm{GL}(V)$ and a pure tensor $v_1\otimes v_2\otimes \cdots \otimes v_n\in V^{\otimes n}$, $$ g \cdot (v_1 \otimes v_2 \otimes \cdots \otimes v_n) := (g v_1) \otimes (g v_2) \otimes \cdots \otimes (g v_n). $$ This extends $\mathbb{C}$-linearly to all of $V^{\otimes n}$. The action defines a representation $\rho :\mathrm{GL}(V)\to \mathrm{GL}(V^{\otimes n})$.

Definition (place-permutation $S_n$-action). For $\sigma \in S_n$, $$ \sigma \cdot (v_1 \otimes v_2 \otimes \cdots \otimes v_n) := v_{\sigma^{-1}(1)} \otimes v_{\sigma^{-1}(2)} \otimes \cdots \otimes v_{\sigma^{-1}(n)}. $$ This defines a representation $\pi :S_n\to \mathrm{GL}(V^{\otimes n})$. The two actions commute: $\rho (g)\pi (\sigma )=\pi (\sigma )\rho (g)$ for all $g\in \mathrm{GL}(V)$, $\sigma \in S_n$, since the $\mathrm{GL}$-action treats every factor identically.

Definition (Weyl module / Schur functor). For a partition $\lambda ⊢n$ with $\ell (\lambda )\le d=\dim V$, the Weyl module (equivalently, the Schur functor applied to $V$) is the $\mathrm{GL}(V)$-representation $$ \mathbb{S}^\lambda V := \mathrm{Hom}{S_n}(S^\lambda, V^{\otimes n}), $$ where $S^{\lambda }$ is the Specht module 07.05.03 of shape $\lambda$. Equivalently, ${\mathbb{S}}^{\lambda }V$ is the $\lambda$-isotypic component of $V^{\otimes n}$ regarded as an $S_n$-module, separated out by the Young symmetriser $c\lambda \in \mathbb{C}[S_n]$: $$ \mathbb{S}^\lambda V = c_\lambda \cdot V^{\otimes n}. $$ The first form makes the $\mathrm{GL}(V)$-action manifest; the second form gives an explicit construction via projection.

Definition (Young symmetriser). For a tableau $T$ of shape $\lambda$, let $R(T)$ and $C(T)$ be the row and column stabilisers in $S_n$. The Young symmetriser is $$ c_T := \Big(\sum_{\sigma \in R(T)} \sigma\Big) \cdot \Big(\sum_{\tau \in C(T)} \mathrm{sgn}(\tau) \tau\Big) \in \mathbb{C}[S_n]. $$ Up to a scalar, $c_T^2=(n!/\dim S^{\lambda })\cdot c_T$, so a suitably normalised $c_T$ is an idempotent projector onto a copy of the Specht module $S^{\lambda }$. The image $c_T\cdot V^{\otimes n}$ is the Weyl module ${\mathbb{S}}^{\lambda }V$.

Counterexamples to common slips

• The bound $\ell (\lambda )\le d$ is sharp: if $\ell (\lambda )>\dim V$, then ${\mathbb{S}}^{\lambda }V=0$. The reason is that the column antisymmetriser kills any tensor with a repeated factor, and if a column has more entries than $d$, every filling forces a repeat.
• The action of $S_n$ on $V^{\otimes n}$ is by place permutation, not by acting on labels. With place permutation, the inverse appears in the formula above so that the action is a left action.
• The decomposition holds in characteristic $0$. In characteristic $p$, the double commutant argument fails (Wedderburn-Artin needs semisimplicity), and the duality is replaced by Schur algebra theory and Green's polynomial representations ^{[Green 1980 LNM 830]}.

Key theorem with proof [Intermediate+]

Theorem (Schur-Weyl duality; Schur 1901, Weyl 1939). Let $V$ be a finite-dimensional complex vector space of dimension $d$, and let $n\ge 1$. Under the commuting actions of $\mathrm{GL}(V)$ and $S_n$, the tensor power decomposes as a $(\mathrm{GL}(V)\times S_n)$-module as $$ V^{\otimes n} = \bigoplus_{\substack{\lambda \vdash n \ \ell(\lambda) \leq d}} \mathbb{S}^\lambda V \otimes S^\lambda, $$ where the sum is over partitions $\lambda$ of $n$ with at most $d$ parts, ${\mathbb{S}}^{\lambda }V$ is the irreducible $\mathrm{GL}(V)$-representation with highest weight $\lambda$ (the Weyl module), and $S^{\lambda }$ is the irreducible $S_n$-representation (the Specht module of shape $\lambda$). Each ${\mathbb{S}}^{\lambda }V$ is non-zero exactly when $\ell (\lambda )\le d$.

The matching is bijective: distinct partitions give distinct irreducible bimodule summands, and every irreducible bimodule summand of $V^{\otimes n}$ has this form.

Proof. The argument is the double commutant theorem applied to the pair of subalgebras of $\mathrm{End}(V^{\otimes n})$ generated by the two actions. Write $$ A := \rho(\mathbb{C}[\mathrm{GL}(V)]) \subseteq \mathrm{End}(V^{\otimes n}), \qquad B := \pi(\mathbb{C}[S_n]) \subseteq \mathrm{End}(V^{\otimes n}), $$ for the subalgebras spanned by the two actions.

Step 1 ($B=A^{\prime }$). The centraliser of $A$ in $\mathrm{End}(V^{\otimes n})$ consists of all endomorphisms commuting with the diagonal $\mathrm{GL}(V)$-action. By a classical theorem of Schur (his 1901 thesis result), the algebra of $\mathrm{GL}(V)$-equivariant endomorphisms of $V^{\otimes n}$ is exactly the algebra of symmetric tensors of $\mathrm{End}(V{)}^{\otimes n}$, generated by the place-permutation action of $S_n$. The argument: the polynomial functions on $\mathrm{End}(V)$ that are $\mathrm{GL}(V)$-invariant under the conjugation $\mathrm{GL}(V)\times \mathrm{End}(V{)}^{\otimes n}\to \mathrm{End}(V{)}^{\otimes n}$ given by simultaneous conjugation are generated by traces of words, which after polarisation reduce to symmetric combinations of the place-permutation operators. Hence $B=A^{\prime }:=\{x\in \mathrm{End}(V^{\otimes n}):xa=ax\text{ for all }a\in A\}$.

Step 2 ($A=B^{\prime }$). The reverse statement is symmetric. The endomorphisms commuting with the place-permutation $S_n$-action on $V^{\otimes n}$ are exactly the diagonal $\mathrm{GL}(V)$-action (extended to its linear span). This is the easier direction: any operator commuting with every $\sigma \in S_n$ must preserve the symmetric-tensor decomposition $V^{\otimes n}={⨁}_{\lambda }{V}^{\otimes n}[\lambda ]$, and on each isotypic component the only $S_n$-equivariant endomorphisms commuting with diagonal $\mathrm{GL}(V)$ are the diagonal-action operators themselves. Equivalently: $A$ is the image of the group algebra of $\mathrm{GL}(V)$, which by Burnside's density theorem maps surjectively onto ${\mathrm{End}}_{S_n}(V^{\otimes n})$ (the centraliser of $B$). Hence $A=B^{\prime }$.

Step 3 (Wedderburn-Artin / double commutant). Both algebras $A$ and $B$ are semisimple (over $\mathbb{C}$): $B$ is the image of the semisimple group algebra $\mathbb{C}[S_n]$, and $A$ is the image of $\mathbb{C}[\mathrm{GL}(V)]$ on the algebraic representation $V^{\otimes n}$, which by complete reducibility of polynomial $\mathrm{GL}(V)$-representations is semisimple. The double commutant theorem (a corollary of Wedderburn-Artin for semisimple algebras) then gives a bimodule decomposition $$ V^{\otimes n} = \bigoplus_i U_i \otimes W_i, $$ where $\{U_i\}$ are the distinct irreducible $A$-modules appearing in $V^{\otimes n}$, $\{W_i\}$ are the distinct irreducible $B$-modules appearing in $V^{\otimes n}$, and $U_i\leftrightarrow W_i$ is a bijection between the two index sets.

Step 4 (identifying the bijection). The irreducible $B$-modules appearing in $V^{\otimes n}$ are exactly the Specht modules $S^{\lambda }$ for partitions $\lambda ⊢n$ with $\ell (\lambda )\le d$; the cut-off comes from the fact that a Specht module $S^{\lambda }$ with $\ell (\lambda )>d$ has its corresponding Young symmetriser annihilating $V^{\otimes n}$ (any column of length exceeding $\dim V$ forces an antisymmetric repeat among $d$-dimensional vectors, which vanishes). Define ${\mathbb{S}}^{\lambda }V$ to be the irreducible $\mathrm{GL}(V)$-summand paired with $S^{\lambda }$ under the bimodule decomposition: ${\mathbb{S}}^{\lambda }V:={\mathrm{Hom}}_{S_n}(S^{\lambda },V^{\otimes n})$. Then by the double commutant decomposition, $$ V^{\otimes n} = \bigoplus_{\substack{\lambda \vdash n \ \ell(\lambda) \leq d}} \mathbb{S}^\lambda V \otimes S^\lambda, $$ and the matching $\lambda \leftrightarrow ({\mathbb{S}}^{\lambda }V,S^{\lambda })$ is bijective by construction.

Step 5 (irreducibility of ${\mathbb{S}}^{\lambda }V$). The Weyl module ${\mathbb{S}}^{\lambda }V$ is irreducible as a $\mathrm{GL}(V)$-representation because the bimodule decomposition pairs distinct irreducibles on each side: ${\mathrm{Hom}}_{\mathrm{GL}(V)\times S_n}({\mathbb{S}}^{\lambda }V\otimes S^{\lambda },{\mathbb{S}}^{\mu }V\otimes S^{\mu })=0$ for $\lambda \ne \mu$, and the scalar-only $\mathrm{GL}(V)\times S_n$-endomorphism algebra forces ${\mathbb{S}}^{\lambda }V$ irreducible by Schur's lemma 07.01.02 applied to each factor.

This completes the proof. $\square$

Bridge. Schur-Weyl duality builds toward the structure of every polynomial representation of $\mathrm{GL}(V)$ and identifies the character theory of $\mathrm{GL}(V)$ with the character theory of $S_n$. The foundational reason it holds is that the diagonal action of $\mathrm{GL}(V)$ on $V^{\otimes n}$ and the place-permutation action of $S_n$ are dual to each other in the precise sense of the double commutant theorem. This is exactly the same pattern that appears again in 07.06.11 (representations of ${\mathfrak{sl}}_2$), where the simplest case $d=2$ identifies the Weyl modules with the symmetric powers ${\mathrm{Sym}}^{n}V$ — the irreducible representations of ${\mathfrak{sl}}_2(\mathbb{C})$. The central insight is that the bimodule decomposition of $V^{\otimes n}$ identifies $\mathrm{GL}(V)$-irreducibles with $S_n$-irreducibles via partitions of $n$, and the Schur functor $V\mapsto {\mathbb{S}}^{\lambda }V$ is the natural-transformation packaging of this identification. The bridge is the recognition that polynomial functors on the category of vector spaces — Schur functors — generalise to a complete description of polynomial $\mathrm{GL}(V)$-representations, with character theory governed by the Schur polynomials $s_{\lambda }$. Putting these together, the Frobenius character formula identifies the symmetric-group characters with the Schur polynomials, and the Weyl character formula for $\mathrm{GL}(V)$ then follows as a corollary of the duality.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Mathlib has RepresentationTheory.Basic and FDRep, plus permutation infrastructure in GroupTheory.Perm.Basic, but no named Schur-Weyl duality. The intended formalisation reads schematically:

import Mathlib.RepresentationTheory.Basic
import Mathlib.RepresentationTheory.FdRep
import Mathlib.GroupTheory.Perm.Basic
import Mathlib.LinearAlgebra.TensorPower
import Mathlib.Combinatorics.YoungDiagram

/-- The diagonal action of GL(V) on V^{⊗n}. -/
noncomputable def diagonalGLAction (V : Type*) [AddCommGroup V] [Module ℂ V]
    [FiniteDimensional ℂ V] (n : ℕ) :
    Representation ℂ (V →ₗ[ℂ] V)ˣ (TensorPower ℂ n V) :=
  sorry

/-- The place-permutation action of S_n on V^{⊗n}. -/
noncomputable def permutationSnAction (V : Type*) [AddCommGroup V] [Module ℂ V]
    (n : ℕ) :
    Representation ℂ (Equiv.Perm (Fin n)) (TensorPower ℂ n V) :=
  sorry

/-- Schur-Weyl duality: V^{⊗n} decomposes as a (GL(V) × S_n)-bimodule
into a direct sum over partitions λ ⊢ n with ℓ(λ) ≤ dim V of
S^λ V ⊗ Specht λ. -/
theorem schurWeylDuality (V : Type*) [AddCommGroup V] [Module ℂ V]
    [FiniteDimensional ℂ V] (n : ℕ) :
    (TensorPower ℂ n V) ≃ₗ[ℂ]
      DirectSum {λ : YoungDiagram // λ.partOf = n ∧ λ.length ≤ FiniteDimensional.finrank ℂ V}
        (fun λ => WeylModule V λ.val ⊗[ℂ] SpechtModule ℂ λ.val) :=
  sorry  -- double commutant theorem applied to GL(V) and S_n

The proof gap is substantive. Mathlib needs the diagonal $\mathrm{GL}(V)$-action on tensor powers as a structured representation, the place-permutation $S_n$-action with explicit commutation relations, the double commutant theorem for semisimple algebras (partially present via Mathlib.RingTheory.SimpleModule), the Weyl module / Schur functor ${\mathbb{S}}^{\lambda }V$ as a named object, and the bimodule decomposition. Each piece is formalisable from existing infrastructure but has not been packaged. The Wedderburn-Artin route requires semisimplicity of the relevant group algebras in characteristic zero, available in Mathlib.RepresentationTheory.Maschke.

Advanced results [Master]

Theorem (Schur-Weyl in the stable range; $n\le d$). When $n\le d=\dim V$, the place-permutation map $\pi :\mathbb{C}[S_n]\to {\mathrm{End}}_{\mathrm{GL}(V)}(V^{\otimes n})$ is an algebra isomorphism, and every partition $\lambda ⊢n$ contributes a non-zero summand ${\mathbb{S}}^{\lambda }V\otimes S^{\lambda }$ to the Schur-Weyl decomposition. The total number of summands is $p(n)$, the number of partitions of $n$.

In the stable range, the duality is at its sharpest: no length-bound cuts off any partition, and the $S_n$-side and $\mathrm{GL}(V)$-side are in perfect dual correspondence. For $n>d$, partitions $\lambda$ with $\ell (\lambda )>d$ are excised on the $\mathrm{GL}(V)$-side (the corresponding ${\mathbb{S}}^{\lambda }V=0$), and the map $\pi$ acquires a kernel of size ${\sum }_{\ell (\lambda )>d}(\dim S^{\lambda }{)}^2$.

Theorem (Frobenius character formula; Frobenius 1900 Sitzungsberichte Berlin). Let $\sigma \in S_n$ have cycle type $\mu =(1^{m_1}{2}^{m_2}\cdots )$. The character of the Specht module $S^{\lambda }$ is $$ \chi^\lambda(\sigma) = \sum_{w \in S_d} \mathrm{sgn}(w) \cdot [\text{coefficient of } x_1^{\lambda_1 + d - 1} x_2^{\lambda_2 + d - 2} \cdots x_d^{\lambda_d}] \prod_{k} p_k(x_1, \ldots, x_d)^{m_k}, $$ where $p_k={\sum }_i{x}_i^k$ is the $k$-th power-sum, and $d\ge \ell (\lambda )$ is any sufficiently large parameter. Equivalently, the symmetric polynomial identity ${\sum }_{\lambda ⊢n}{\chi }^{\lambda }(\mu )s_{\lambda }=p_{\mu }$ holds in the ring of symmetric polynomials.

The Frobenius formula is the character-theoretic content of Schur-Weyl: the symmetric-group characters and the $\mathrm{GL}(V)$-characters (Schur polynomials) are dual under the standard inner product on the ring of symmetric functions. The transition matrix between the bases $\{s_{\lambda }\}$ and $\{p_{\mu }\}$ is the character table of $S_n$.

Theorem (Weyl character formula for $\mathrm{GL}(V)$; Weyl 1925-26 Mathematische Zeitschrift). For $\lambda$ a dominant weight of $\mathrm{GL}(V)$ (a partition with $\ell (\lambda )\le d$), the character of ${\mathbb{S}}^{\lambda }V$ at $g\in \mathrm{GL}(V)$ with eigenvalues $x_1,\dots ,x_d$ is the Schur polynomial $$ \chi_{\mathbb{S}^\lambda V}(g) = s_\lambda(x_1, \ldots, x_d) = \frac{\det\big(x_i^{\lambda_j + d - j}\big){i,j}}{\det\big(x_i^{d - j}\big){i,j}}. $$ This is the Weyl character formula for the general linear group, derived in 1925-26 by the unitarian trick from the representation theory of the compact unitary group $\mathrm{U}(d)$.

Schur-Weyl gives the formula via the Frobenius-character side: $s_{\lambda }$ is the character of ${\mathbb{S}}^{\lambda }V$ by the duality, and the Jacobi-Trudi alternant representation of $s_{\lambda }$ is the Weyl character formula in its determinantal form.

Theorem (Schur functors as polynomial functors; Schur 1901 dissertation, modern form Macdonald 1995 §I.7). The assignment $V\mapsto {\mathbb{S}}^{\lambda }V$ is a polynomial functor from finite-dimensional complex vector spaces to themselves, homogeneous of degree $n=|\lambda |$. The composition of Schur functors is governed by the Littlewood-Richardson coefficients: $$ \mathbb{S}^\lambda V \otimes \mathbb{S}^\mu V = \bigoplus_\nu c^\nu_{\lambda \mu} \cdot \mathbb{S}^\nu V, $$ and the Schur polynomials $s_{\lambda }$ form a $\mathbb{Z}$-basis of the ring $\Lambda$ of symmetric functions, multiplying as $s_{\lambda }\cdot s_{\mu }={\sum }_{\nu }{c}_{\lambda \mu }^{\nu }{s}_{\nu }$.

The polynomial-functor formulation extends Schur-Weyl from $\mathrm{GL}(V)$ to the categorical setting: Schur functors are the universal polynomial functors on vector spaces, and any natural transformation between polynomial functors of the same degree factors through the Schur decomposition.

Theorem (${\mathfrak{sl}}_2$ case as smallest Schur-Weyl). For $V={\mathbb{C}}^2$, Schur-Weyl gives $$ V^{\otimes n} = \bigoplus_{k = 0}^{\lfloor n/2 \rfloor} \mathrm{Sym}^{n - 2k} V \otimes S^{(n-k, k)} $$ where ${\mathrm{Sym}}^{n-2k}V$ is the irreducible ${\mathfrak{sl}}_2$-representation of highest weight $n-2k$. The decomposition is the algebraic version of angular-momentum addition in physics: adding $n$ spin-$1/2$ particles gives total spins $n/2,n/2-1,\dots ,0$ or $1/2$, with multiplicities counted by $\dim S^{(n-k,k)}$.

The ${\mathfrak{sl}}_2$ case is the simplest substantive example of Schur-Weyl and the input to the angular-momentum coupling rules in quantum mechanics. The Clebsch-Gordan coefficients on the physics side are the matrix elements of the Schur-Weyl isomorphism.

Theorem (centraliser-algebra interpretation). The Schur algebra $S_{\mathbb{C}}(d,n):={\mathrm{End}}_{S_n}(V^{\otimes n})$ where $V={\mathbb{C}}^d$ has the same finite-dimensional representation theory as polynomial ${\mathrm{GL}}_d$-representations of degree $n$. Green's monograph (1980 LNM 830) treats Schur-Weyl from this algebra-theoretic side and extends the theory to arbitrary infinite fields and to quantum groups.

The Schur algebra is the algebraic skeleton of the polynomial representation theory of ${\mathrm{GL}}_d$. Over a field of positive characteristic, Schur-Weyl in its bimodule form fails (Wedderburn-Artin requires semisimplicity), but the Schur algebra $S_k(d,n)$ retains a rich representation theory described by Green's quasi-hereditary algebra structure, with implications for modular representation theory.

Synthesis. Schur-Weyl duality is the foundational reason every polynomial representation of $\mathrm{GL}(V)$ is built from the symmetric-group representations via Schur functors. The central insight is that the two natural actions on $V^{\otimes n}$ — diagonal $\mathrm{GL}(V)$ and place-permutation $S_n$ — are dual to each other in the precise sense of the double commutant theorem, with the duality indexed by partitions of $n$. This is exactly the same pattern that appears again in 07.06.11 (representations of ${\mathfrak{sl}}_2$), where the simplest case $\dim V=2$ recovers the angular-momentum decomposition that physicists call "adding spin-1/2 particles" and that mathematicians call "Clebsch-Gordan coupling". Putting these together, the duality identifies the character theory of $S_n$ (Frobenius's formula) with the character theory of $\mathrm{GL}(V)$ (Weyl's formula via Jacobi-Trudi alternants); the bridge is the symmetric-function ring $\Lambda$, whose Schur basis $s_{\lambda }$ encodes both sides.

The duality generalises in three directions. Replacing $\mathrm{GL}(V)$ with the orthogonal or symplectic group, and $S_n$ with the Brauer algebra, gives the Brauer-Weyl duality of Brauer 1937 Annals of Mathematics. Replacing the field $\mathbb{C}$ with a positive-characteristic field, and the group algebra with the Schur algebra, gives Green's modular Schur-Weyl. Replacing $\mathrm{GL}(V)$ with the quantum group $U_q({\mathfrak{gl}}_n)$ and $S_n$ with the Hecke algebra $H_q(S_n)$ gives the Jimbo-Drinfeld quantum Schur-Weyl duality, which is dual to itself in the precise sense that the duality identifies the centraliser of one quantum group with the other. The same skeleton — two algebras acting on a tensor power, each one the centraliser of the other, with irreducibles paired by a combinatorial index set — recurs through every flavour of representation theory and is what justifies treating "Schur-Weyl-type duality" as a structural principle rather than a single theorem.

Full proof set [Master]

Theorem (Schur-Weyl duality), proof. Given in the Intermediate-tier section: double commutant theorem applied to the pair $(A,B)=(\rho (\mathbb{C}[\mathrm{GL}(V)]),\pi (\mathbb{C}[S_n]))$ acting on $V^{\otimes n}$. The two-sided centraliser identifications $A=B^{\prime }$ and $B=A^{\prime }$ come from Schur's 1901 result on $\mathrm{GL}(V)$-equivariant tensor endomorphisms (Step 1) and the Burnside density theorem (Step 2). Wedderburn-Artin for semisimple algebras over $\mathbb{C}$ then yields the bimodule decomposition. The partition $\lambda \leftrightarrow ({\mathbb{S}}^{\lambda }V,S^{\lambda })$ matching is bijective by construction, with the length-bound $\ell (\lambda )\le d$ coming from the vanishing of the Young symmetriser on tensors of "too few" factors. $\square$

Proposition (Schur-Weyl in the stable range, $n\le d$), proof. When $n\le d$, every partition $\lambda ⊢n$ has $\ell (\lambda )\le n\le d$, so the length-bound is automatic and every partition contributes. The annihilator of $V^{\otimes n}$ in $\mathbb{C}[S_n]$ — namely ${⨁}_{\ell (\lambda )>d}\mathrm{End}(S^{\lambda })$ — is the zero subalgebra. Hence $\pi :\mathbb{C}[S_n]\to {\mathrm{End}}_{\mathrm{GL}(V)}(V^{\otimes n})$ is injective, and by Schur-Weyl surjective, hence an algebra isomorphism. The total summand count is $p(n)$, the number of partitions of $n$. $\square$

Proposition (Frobenius character formula), proof sketch. Compute the bivariate character $\mathrm{tr}(\rho (g)\pi (\sigma ))$ on $V^{\otimes n}$ in two ways. By Schur-Weyl, it equals ${\sum }_{\lambda }{\chi }_{{\mathbb{S}}^{\lambda }V}(g)\cdot {\chi }^{\lambda }(\sigma )={\sum }_{\lambda }{s}_{\lambda }(x_1,\dots ,x_d){\chi }^{\lambda }(\sigma )$, where $x_1,\dots ,x_d$ are eigenvalues of $g$. By direct computation, for $\sigma$ of cycle type $\mu$, the trace equals $p_{\mu }(x_1,\dots ,x_d)={\prod }_k{p}_k^{m_k}$, where the factor $p_k$ records the trace of $g^k$ on each $k$-cycle of $\sigma$ acting on $V^{\otimes k}$. Equating: ${\sum }_{\lambda }{\chi }^{\lambda }(\mu )s_{\lambda }=p_{\mu }$. Inverting via Schur-polynomial orthogonality gives the dual formula $s_{\lambda }={\sum }_{\mu }{z}_{\mu }^{-1}{\chi }^{\lambda }(\mu )p_{\mu }$. Both sides live in the ring of symmetric functions $\Lambda$ in infinitely many variables. $\square$

Proposition (Weyl character formula for $\mathrm{GL}(V)$), proof. By Schur-Weyl, the character of ${\mathbb{S}}^{\lambda }V$ at $g$ with eigenvalues $x_1,\dots ,x_d$ is the Schur polynomial $s_{\lambda }(x_1,\dots ,x_d)$. The Jacobi-Trudi identity expresses $$ s_\lambda(x_1, \ldots, x_d) = \frac{\det(x_i^{\lambda_j + d - j}){i, j}}{\det(x_i^{d-j}){i, j}}. $$ The numerator is the alternant $a_{\lambda +\delta }$, where $\delta =(d-1,\dots ,1,0)$ is the half-sum of positive roots of $\mathrm{GL}(V)$; the denominator is the Vandermonde ${\prod }_{i<j}(x_i-x_j)=a_{\delta }$. Recognising the numerator as ${\sum }_{w\in S_d}\mathrm{sgn}(w)x^{w(\lambda +\delta )}$ and the denominator as the Weyl denominator gives the Weyl character formula in its standard form. $\square$

Proposition (Schur functor decomposition), proof. The polynomial-functor structure is built directly into the construction: ${\mathbb{S}}^{\lambda }$ is defined on a vector space $V$ by ${\mathbb{S}}^{\lambda }V=c_{\lambda }\cdot V^{\otimes n}$ for the Young symmetriser $c_{\lambda }$, and this assignment extends to morphisms by functoriality of tensor powers. Polynomial-functor-ness follows because ${\mathbb{S}}^{\lambda }(V)\subseteq V^{\otimes n}$ as a sub-functor; tensor power is polynomial of degree $n$. For the Littlewood-Richardson decomposition of ${\mathbb{S}}^{\lambda }V\otimes {\mathbb{S}}^{\mu }V$: applying Schur-Weyl to the tensor power $V^{\otimes (n+m)}$ with $n=|\lambda |,m=|\mu |$, and selecting the $(\lambda ,\mu )$-isotypic component under the embedding $S_n\times S_m\hookrightarrow S_{n+m}$, gives the formula $$ \mathrm{Ind}{S_n \times S_m}^{S{n+m}}(S^\lambda \boxtimes S^\mu) = \bigoplus_\nu c^\nu_{\lambda \mu} S^\nu, $$ where $c_{\lambda \mu }^{\nu }$ are the Littlewood-Richardson coefficients. Dualising under Schur-Weyl translates this to the Schur-functor side. $\square$

Proposition (${\mathfrak{sl}}_2$ case decomposition), proof. For $V={\mathbb{C}}^2$, the partitions of $n$ with at most $2$ parts are $(n-k,k)$ for $0\le k\le \lfloor n/2\rfloor$. The Weyl module ${\mathbb{S}}^{(n-k,k)}{\mathbb{C}}^2$ has highest weight $\lambda =(n-k)-k=n-2k$ in the standard ${\mathfrak{sl}}_2$-convention, with dimension $n-2k+1$. The Specht module $S^{(n-k,k)}$ has dimension $\left(\genfrac{}{}{0ex}{}{n}{k}\right)-\left(\genfrac{}{}{0ex}{}{n}{k-1}\right)$ by the hook length formula 07.05.02. The total dimension check: $$ \sum_{k=0}^{\lfloor n/2 \rfloor} (n - 2k + 1) \left[\binom{n}{k} - \binom{n}{k-1}\right] = 2^n, $$ which can be verified by a generating-function computation or by recognising the left side as the dimension of $V^{\otimes n}=({\mathbb{C}}^2{)}^{\otimes n}={\mathbb{C}}^{2^n}$. The physical interpretation: $n$ spin-$1/2$ particles couple to total spin $j=n/2-k$ for $k=0,1,\dots ,\lfloor n/2\rfloor$, with multiplicity $\dim S^{(n-k,k)}$ given by the corresponding Catalan-like count. $\square$

Proposition (centraliser-algebra interpretation), stated without proof — see Green 1980 Polynomial Representations of ${\mathrm{GL}}_n$ LNM 830 ^[pending]. The Schur algebra $S_{\mathbb{C}}(d,n)={\mathrm{End}}_{S_n}(V^{\otimes n})$ has finite-dimensional category of modules equivalent to the polynomial ${\mathrm{GL}}_d$-representations of degree $n$. The argument is Morita-theoretic: $V^{\otimes n}$ is a finitely generated projective generator for $S_{\mathbb{C}}(d,n)$, and the functor $M\mapsto V^{\otimes n}{\otimes }_{S_{\mathbb{C}}(d,n)}M$ is an equivalence of categories. Over a positive-characteristic field $k$, the Schur algebra $S_k(d,n)$ replaces Schur-Weyl in its bimodule form, retaining a rich representation theory but without the double commutant decomposition. The proof in Green's monograph extends to quantum Schur-Weyl, with $S_k(d,n)$ replaced by the $q$-Schur algebra and $\mathbb{C}[S_n]$ by the Hecke algebra. $\square$

Connections [Master]

• Symmetric group representation 07.05.01. Schur-Weyl duality matches the irreducible representations of $\mathrm{GL}(V)$ with the irreducible representations of $S_n$ via partitions $\lambda ⊢n$. The $S_n$-side is exactly the foundational classification of 07.05.01: irreducibles indexed by partitions, with characters governed by the Frobenius formula. Schur-Weyl is the bridge that exports this classification to the general-linear group.

• Specht module 07.05.03. The $S_n$-irreducible $S^{\lambda }$ in the bimodule decomposition is the Specht module of 07.05.03. The Schur functor ${\mathbb{S}}^{\lambda }V$ is constructed as ${\mathrm{Hom}}_{S_n}(S^{\lambda },V^{\otimes n})$, which makes the Specht module the structural input on the symmetric-group side of the duality. The Young symmetriser $c_{\lambda }\in \mathbb{C}[S_n]$ that projects onto $S^{\lambda }$ is the same operator that cuts out ${\mathbb{S}}^{\lambda }V$ from $V^{\otimes n}$.

• Representations of ${\mathfrak{sl}}_2(\mathbb{C})$ 07.06.11. The simplest case of Schur-Weyl duality is $V={\mathbb{C}}^2$, where $\mathrm{GL}(V)=\mathrm{GL}(2,\mathbb{C})$ and the Weyl modules are the symmetric powers ${\mathrm{Sym}}^{n}V$ — the irreducible representations of ${\mathfrak{sl}}_2$. The Schur-Weyl decomposition $V^{\otimes n}={⨁}_k{\mathrm{Sym}}^{n-2k}V\otimes S^{(n-k,k)}$ is the algebraic content of the Clebsch-Gordan coupling rule for adding spin-1/2 systems in quantum mechanics.

• Representations of ${\mathfrak{sl}}_3(\mathbb{C})$ 07.06.12. Taking $V={\mathbb{C}}^3$ and $\mathrm{GL}(V)=\mathrm{GL}(3,\mathbb{C})$, Schur-Weyl duality is the route by which the Fulton-Harris framework derives the ${\mathfrak{sl}}_3$ irreducibles $V_{a,b}$ from the symmetric-group side: the Weyl module ${\mathbb{S}}^{\lambda }{\mathbb{C}}^3$ for $\lambda =({\lambda }_1,{\lambda }_2,{\lambda }_3)$ with $\ell (\lambda )\le 3$ is the irreducible $\mathrm{GL}(3,\mathbb{C})$-representation of highest weight $\lambda -({\lambda }_3,{\lambda }_3,{\lambda }_3)$, which descends to the ${\mathfrak{sl}}_3$-irreducible $V_{a,b}$ with $a={\lambda }_1-{\lambda }_2$, $b={\lambda }_2-{\lambda }_3$. The hexagonal weight diagram of $V_{a,b}$ at 07.06.12 is therefore the Schur-functor image of the partition $(a+b,b,0)$, and the Weyl dimension formula $\dim V_{a,b}=(a+1)(b+1)(a+b+2)/2$ is the Schur polynomial $s_{(a+b,b,0)}(x_1,x_2,x_3)$ evaluated at $x_i=1$. Schur-Weyl is the foundational bridge that exports the partition combinatorics of $S_n$ to the rank-two highest-weight classification.

• Tensor product of representations 07.01.06. Schur-Weyl is a statement about the structure of the tensor power $V^{\otimes n}$, the iterated tensor product of 07.01.06. The duality identifies the structure of this iterated tensor as a $(\mathrm{GL}(V)\times S_n)$-bimodule and extracts polynomial $\mathrm{GL}(V)$-representations as factors. The Littlewood-Richardson rule extends this to tensor products of Schur functors, ${\mathbb{S}}^{\lambda }V\otimes {\mathbb{S}}^{\mu }V$.

• Schur's lemma 07.01.02. The irreducibility of each Schur functor ${\mathbb{S}}^{\lambda }V$ as a $\mathrm{GL}(V)$-representation follows from Schur's lemma applied to the bimodule structure: distinct $(\lambda ,\mu )$-isotypic components have no inter-twiners, forcing ${\mathbb{S}}^{\lambda }V$ irreducible. The lemma is invoked at the final step of the double-commutant proof to separate the Wedderburn matrix-algebra summands into irreducible factors.

• Weyl character formula 07.06.07. The character of ${\mathbb{S}}^{\lambda }V$ as a $\mathrm{GL}(V)$-representation is the Schur polynomial $s_{\lambda }(x_1,\dots ,x_d)$. The Jacobi-Trudi expression of $s_{\lambda }$ as a ratio of alternants is the Weyl character formula for $\mathrm{GL}(V)$, and Schur-Weyl duality provides one of the routes (originally Weyl's 1925-26 derivation) for proving it. The Weyl-character-formula unit 07.06.07 gives the abstract Lie-algebraic form; Schur-Weyl gives the concrete-tensor-power realisation.

Historical & philosophical context [Master]

Issai Schur's 1901 Berlin dissertation Über eine Klasse von Matrizen, die sich einer gegebenen Matrix zuordnen lassen introduced the algebra of $\mathrm{GL}(V)$-equivariant tensor endomorphisms, identified it with the symmetric-tensor algebra of $\mathrm{End}(V{)}^{\otimes n}$, and proved the surjectivity of $\mathbb{C}[S_n]\to {\mathrm{End}}_{\mathrm{GL}(V)}(V^{\otimes n})$ ^[pending]. The 1901 work was the originator of the duality phenomenon, though Schur did not state it in the bimodule-decomposition form that is now standard. Schur's PhD advisor at Berlin was Frobenius, whose 1900 paper Über die Charaktere der symmetrischen Gruppe (Sitzungsberichte der Königlich Preussischen Akademie der Wissenschaften zu Berlin 1900, 516-534) ^[pending] had established the character theory of $S_n$ via the Frobenius character formula. The 1901 thesis was Schur's translation of his teacher's symmetric-group character theory into the parallel setting of $\mathrm{GL}(V)$.

Hermann Weyl's 1925-26 papers in Mathematische Zeitschrift (Theorie der Darstellung kontinuierlicher halbeinfacher Gruppen durch lineare Transformationen, Math. Z. 23 (1925), 271-309; 24 (1926), 328-376, 377-395, 789-791) ^[pending] introduced the unitarian trick and derived the Weyl character formula for compact Lie groups. The corresponding character formula for $\mathrm{GL}(V)$ followed by passage from the compact $\mathrm{U}(d)$ to $\mathrm{GL}(d,\mathbb{C})$ via algebraic continuation. Weyl's 1939 monograph The Classical Groups: Their Invariants and Representations (Princeton University Press) ^[pending] consolidated the Schur-Weyl story into its modern form, naming the duality and presenting the bimodule decomposition explicitly. The duality is named for Schur and Weyl because of these complementary contributions: Schur 1901 for the centraliser algebra side, Weyl 1939 for the bimodule formulation.

The modern symmetric-function side was developed by Alfred Young (1900-1933 series in Proceedings of the London Mathematical Society) ^[pending], who built the explicit projectors $c_{\lambda }$ now called Young symmetrisers; by Wilhelm Specht (1935 Mathematische Zeitschrift) ^[pending], who gave the modular-friendly construction of $S_n$-irreducibles; and by Donald Knutson (1973 $\lambda$-rings and the Representation Theory of the Symmetric Group) ^[pending], who packaged the polynomial-functor formulation. James 1978 The Representation Theory of the Symmetric Group (LNM 682) ^[pending] gave the canonical modern textbook treatment of the $S_n$-side. Macdonald 1995 Symmetric Functions and Hall Polynomials (Oxford University Press, 2nd ed.) ^[pending] treats the symmetric-function side, with Chapter I §7 devoted to the Schur basis and its duality properties. Green 1980 Polynomial Representations of ${\mathrm{GL}}_n$ (Springer LNM 830) ^[pending] handles the polynomial-representation theory in arbitrary characteristic via the Schur algebra. Quantum Schur-Weyl was introduced by Jimbo 1986 and developed by Drinfeld; Brauer-Weyl duality for orthogonal and symplectic groups dates to Brauer 1937 Annals of Mathematics 38, 857-872 ^[pending].

Bibliography [Master]

@phdthesis{Schur1901,
  author  = {Schur, Issai},
  title   = {{\"U}ber eine Klasse von Matrizen, die sich einer gegebenen Matrix zuordnen lassen},
  school  = {Friedrich-Wilhelms-Universit{\"a}t zu Berlin},
  year    = {1901}
}

@article{Frobenius1900,
  author  = {Frobenius, Ferdinand Georg},
  title   = {{\"U}ber die Charaktere der symmetrischen Gruppe},
  journal = {Sitzungsberichte der K{\"o}niglich Preussischen Akademie der Wissenschaften zu Berlin},
  year    = {1900},
  pages   = {516--534}
}

@article{Weyl1925,
  author  = {Weyl, Hermann},
  title   = {Theorie der Darstellung kontinuierlicher halbeinfacher Gruppen durch lineare Transformationen, I},
  journal = {Mathematische Zeitschrift},
  volume  = {23},
  year    = {1925},
  pages   = {271--309}
}

@article{Weyl1926,
  author  = {Weyl, Hermann},
  title   = {Theorie der Darstellung kontinuierlicher halbeinfacher Gruppen durch lineare Transformationen, II--III},
  journal = {Mathematische Zeitschrift},
  volume  = {24},
  year    = {1926},
  pages   = {328--376, 377--395, 789--791}
}

@book{Weyl1939ClassicalGroups,
  author    = {Weyl, Hermann},
  title     = {The Classical Groups: Their Invariants and Representations},
  publisher = {Princeton University Press},
  year      = {1939}
}

@article{Young19001933,
  author  = {Young, Alfred},
  title   = {On Quantitative Substitutional Analysis, I--IX},
  journal = {Proceedings of the London Mathematical Society},
  year    = {1900--1933}
}

@article{Specht1935,
  author  = {Specht, Wilhelm},
  title   = {Die irreduziblen Darstellungen der symmetrischen Gruppe},
  journal = {Mathematische Zeitschrift},
  volume  = {39},
  year    = {1935},
  pages   = {696--711}
}

@book{James1978,
  author    = {James, Gordon D.},
  title     = {The Representation Theory of the Symmetric Group},
  publisher = {Springer-Verlag},
  series    = {Lecture Notes in Mathematics},
  volume    = {682},
  year      = {1978}
}

@book{Macdonald1995,
  author    = {Macdonald, Ian G.},
  title     = {Symmetric Functions and Hall Polynomials},
  publisher = {Oxford University Press},
  edition   = {2},
  year      = {1995}
}

@book{Green1980,
  author    = {Green, James A.},
  title     = {Polynomial Representations of $\mathrm{GL}_n$},
  publisher = {Springer-Verlag},
  series    = {Lecture Notes in Mathematics},
  volume    = {830},
  year      = {1980}
}

@book{FultonHarrisRepresentationTheory,
  author    = {Fulton, William and Harris, Joe},
  title     = {Representation Theory: A First Course},
  publisher = {Springer-Verlag},
  series    = {Graduate Texts in Mathematics},
  volume    = {129},
  year      = {1991}
}

@book{GoodmanWallach,
  author    = {Goodman, Roe and Wallach, Nolan R.},
  title     = {Symmetry, Representations, and Invariants},
  publisher = {Springer},
  series    = {Graduate Texts in Mathematics},
  volume    = {255},
  year      = {2009}
}

@book{SaganSymmetricGroup,
  author    = {Sagan, Bruce E.},
  title     = {The Symmetric Group: Representations, Combinatorial Algorithms, and Symmetric Functions},
  publisher = {Springer},
  edition   = {2},
  series    = {Graduate Texts in Mathematics},
  volume    = {203},
  year      = {2001}
}

@article{Brauer1937,
  author  = {Brauer, Richard},
  title   = {On Algebras Which are Connected with the Semisimple Continuous Groups},
  journal = {Annals of Mathematics},
  volume  = {38},
  year    = {1937},
  pages   = {857--872}
}

@book{Knutson1973,
  author    = {Knutson, Donald},
  title     = {$\lambda$-rings and the Representation Theory of the Symmetric Group},
  publisher = {Springer-Verlag},
  series    = {Lecture Notes in Mathematics},
  volume    = {308},
  year      = {1973}
}